Optical coherence tomography (OCT) is a method for noncontact optical imaging taking advantage of sequential or scanned distance measurements developed primarily in the 1990's. In biological and biomedical imaging applications, OCT allows for micrometer-scale imaging noninvasively in transparent and translucent biological tissues. The longitudinal ranging capability of OCT is based on low-coherence interferometry, in which light from a broadband source is split between illuminating the sample of interest and a reference path. The interference pattern of light reflected or backscattered from the sample and light from the reference delay contains information about the location and scattering amplitude of the scatterers in the sample. In conventional (time-domain) OCT, this information is extracted by scanning the reference path delay and detecting the resulting interferogram pattern as a function of that delay.
The envelope of the interferogram pattern thus detected represents a map of the reflectivity of the sample versus depth, called an “A-scan”, with depth resolution given by the coherence length of the source. In conventional OCT systems, multiple A-scans are acquired while the sample beam is scanned laterally across the tissue surface, making a continuous series of distance measurements and building up a two-dimensional map of reflectivity versus depth and lateral extent called a “B-scan.” The lateral resolution of the B-scan is given by the confocal resolving power of the sample arm optical system, which is usually given by the size of the focused optical spot in the tissue.
Time-domain OCT systems have been designed to operate at moderate (˜1 image/sec) and high speeds (up to video rate), and have been applied for imaging in biological applications such as imaging of embryonic development, as well as in medical diagnostic applications such as imaging the structures of the anterior and posterior segments of the eye, the skin, the gastrointestinal tract, and other tissues. Specialized probes, endoscopes, catheters, and biomicroscope attachments have been designed to allow for OCT imaging in these applications.
The time-domain approach in conventional OCT has been by far the most successful to date in supporting biological and medical applications, and all in-vivo human clinical trials of OCT to date have utilized this approach. However, the time-domain approach in OCT suffers from some limitations. First, the requirement for mechanical scanning, such as with bulk optics, of the reference delay in conventional OCT introduces complexity, expense, and reduced reliability, especially those which image at high speed and acquire A-scans at kilohertz rates. The mechanical scanning reference delay line is typically the most complex optical apparatus in high-speed conventional OCT systems, and can be quite bulky as well. Second, since conventional OCT images are built up serially using a single detector and collecting one pixel of image information at a time, no advantage is taken of modern 1D and 2D array detection technologies which dominate other forms of optical imaging.
The serial collection or scanning approach of time-domain OCT is also very wasteful of sample arm light, in that an entire column of pixels is illuminated by that light while reflected light is only collected from one pixel at a time. This wastefulness of sample arm light is costly because sources of broadband light suitable for use in OCT systems are typically expensive and limited in their output power capability, and also because optical damage to tissue structures often limits the maximum power which may be used in OCT imaging, particularly in the retina. Where there is a limit on the amount of light which may be used to illuminate the sample, the wastefulness of sample arm light translates directly into increased image acquisition time. Further, the serial scanning approach in conventional OCT requires that the sample under investigation remains stationary during the acquisition of each A-scan, otherwise motion artifacts may appear in the image. Finally, primarily because of the requirement for a mechanical delay scan, conventional high-speed OCT systems are typically expensive, bulky, and require frequent optical alignment.
A potential solution to this need for a new approach has been variously termed spectral radar, Fourier-domain OCT (FDOCT), complex Fourier OCT, Optical Frequency-domain imaging, and swept-source OCT. In FDOCT, a different form of low-coherence interferometry is used in which the reference delay is fixed (except for potential wavelength-scale delay modulation in some implementations), and information about the location and amplitude of scatterers in the sample is derived from the optical spectrum of the light returning from the sample and mixing with the reference. This spectral information is typically acquired by spectrally dispersing the detector arm light using a spectrometer and detecting it with an array detector such as a charge-coupled device (CCD), or else by using a single detector and sweeping the source frequency as a function of time
The A-scan data collected using FDOCT can be shown to be related (see below) to the inverse Fourier transform of the spectral data thus acquired. Initial implementations of FDOCT suffered from image artifacts resulting from: 1) large direct-current (DC) signals appearing on the detector array arising from non-interfering light returning from the reference delay and the sample, thus dwarfing the much smaller interferometric signals; and 2) autocorrelation of light signals between different reflections within the sample. As a result, initial results of FDOCT imaging were filled with artifacts and were not comparable to images obtained with time-domain OCT.
Recently, newer implementations of FDOCT have appeared which take advantage of techniques well known from phase-shifting interferometry (PSI) to eliminate the sources of both of the artifacts mentioned above. Since both artifacts resulted from light appearing on the detector array which does not arise from interference between sample and reference arm light, the recently introduced technique of complex FDOCT eliminates these artifacts by acquiring multiple spectra with different phase shifts introduced into the reference delay path.
In a simple implementation of FDOCT, the reference delay consists of a mirror mounted on a piezoelectric actuator (PZT). One spectrum is acquired at a given position of the mirror, and then another is acquired with a path-length delay of π/2 (resulting in a round-trip phase shift of π) introduced into the reference arm by the PZT. It is straightforward to show that this π phase shift reverses the sign of the interferometric light components but has no effect on the DC components of the detector arm light, so subtracting the spectra obtained at 0 and π phase shifts results in a spectrum free of DC artifacts. This spectrum can be considered the real part of the complex Fourier transform of the A-scan. Thus, taking the inverse Fourier transform reconstructs the original A-scan. However, since only the real part of the complex Fourier spectrum is acquired, the A-scan data reconstructed is restricted to be symmetric. Specifically, f(−z)=f*(z), and thus only A-scan data for positive displacements (i.e., z>0) can be reconstructed.
As a further refinement of this phase-shifting technique, an additional spectrum may be acquired for a path-length delay of π/4 (corresponding to a round-trip phase shift of π/2). This spectrum (also optionally corrected for DC components by division by one of the other spectra or by subtraction with a spectrum acquired with a path-length delay of 3π/4) may be considered the imaginary part of the complex Fourier transform of the A-scan. Thus, taking the inverse Fourier transform of the complete complex spectrum (resulting from all two, three, or four phase measurements) allows for unambiguous reconstruction of all depths in the sample limited only by spatial sampling considerations. Additional refinements to this approach may be applied which are commonplace in phase-shifting interferometry, such as the use of additional phase delays for increased accuracy in measuring the complex spectrum.
Complex FDOCT thus addresses several of the needs for OCT systems with decreased complexity and cost and increased reliability, having a mostly fixed reference delay and utilizing an array detector. However, serious limitations to these prior art complex FDOCT implementations include: 1) a means is still required for displacing the reference delay by distances on the scale of a wavelength; all prior systems perform this function by using bulk optical devices outside of the reference arm optical fiber; and 2) the spectra obtained at different reference phases are obtained sequentially, thus the sample and reference arms must be maintained interferometrically motionless during the entire A-scan spectrum acquisition.